Peri-implantitis: A Complex Condition With Unique Dynamic Characteristics

Methods: Clinical and radiographic variables were selected from a group of 94 implant-treated patients of a private periodontal practice. Data for 340 implants with a mean 7.1± 4.1 years of service were collected. Kernel density estimation (KDE) on patients mean peri-implant bone loss (PMPIBL) dataset was performed. Thirteen demographic, periodontal and prosthetic parameters were evaluated by principal component analysis (PCA) for explaining the variability of PMPIBL. A k-nearest neighbors (KNN) model performed supervised prediction of PMPIBL. Fractal dimensions (FD) of individual implant marginal bone loss (IIMBL) datasets from different jaw bone sites (JBS) were estimated.

Results: KDE provided evidence for two clusters of implant-treated patients at modes of 1.7 mm and 4 mm of the PMPIBL distribution. Five parameters found by PCA as principal (in the order that follows: medical conditions compromising the host immune response and bone metabolism, diabetes, age, cantilevers in prostheses, compliance with recall schedule) were inserted in KNN that predicted PMPIBL with accuracy (sum-of-squared error = 0.03). A low FD indicates increased susceptibility to peri-implantitis and JBS contributes significantly in explaining IIMBL variance (p= 0.032).

Conclusions: This study identified two main groups of implant-treated patients and suggested clinical predictors of susceptibility to peri-implantitis that might enhance our ability to design personalized treatment and maintenance regimes for implant-treated patients.